Is it possible for computers to think? Join a discussion on artificial intelligence. (To participate, sign in. Then click your browser's "Back" button twice and enter the discussion.)
Survey Said: Human
From The Post
On the Web
On IBM's Web Site
Editor's Note: Some of the above links will take you out of The Post's Web site.
To return, use the Back button on your browser.
Go to Technology Post
By Rajiv Chandrasekaran
Washington Post Staff Writer
Monday, May 12, 1997; Page A01
NEW YORK, May 11 -- In a stunning showdown between man and machine, the IBM supercomputer Deep Blue decisively beat world chess champion Garry Kasparov today, the first time a computer has been able to defeat the best human player in a match.
A visibly upset Kasparov stormed out of the small match room after only about an hour of play, effectively resigning the sixth -- and final -- game with a scant 19 moves played. Most chess experts here said Kasparov, who appeared frustrated from the start of today's game, likely would have been conquered by the computer within a few moves.
"This was the single most historic event in the history of chess," said Daniel Edelman, a grandmaster and an editor of the American Chess Journal.
"We have a machine here that is truly remarkable," said David Levy, the vice president of the International Computer Chess Association. "This was an amazing victory."
Kasparov, in a postgame news conference, accused International Business Machines Corp. of building a machine specifically to defeat him. "It was nothing to do about science. . . . It was zeal to beat Garry Kasparov," he said. "And when a big corporation with unlimited resources would like to do so, there are many ways to achieve the result. And the result was achieved."
Kasparov, who had never lost a match until today, said he "cracked under the pressure" of playing the computer. He apologized for his performance and his hasty exit, saying he felt "ashamed by what I did at the end of this match." At the same time, he said his loss "has nothing to do with the computer being unbeatable."
Other computer and chess experts here disagreed, predicting Deep Blue and its progeny will regularly be able to defeat the world's top players. Deep Blue's strong performance surprised many of them, who expected Kasparov to be able to trick the computer by playing unconventional moves.
But exactly the opposite happened. Deep Blue, which can evaluate 200 million possible moves each second, was expected to play a brute-force sort of game, like a tennis player smashing only powerful shots across the court. Instead, the computer dazzled spectators -- and Kasparov himself -- with its ability to develop strategies as a human player would. It was akin to surprising Kasparov with volleys and drop-shots across the chessboard.
To make matters worse for him, Kasparov said his efforts to change his playing style -- sometimes trying to trick the computer and other times substituting his usual aggressive style for a more measured approach -- essentially backfired. "I was playing against myself and something I couldn't recognize," he said.
Deep Blue won the match 3 1/2-2 1/2. Kasparov won the first game, Deep Blue the second, and the two agreed upon draws in the third, fourth and fifth.
Kasparov said he was unable to maintain his concentration today because of his resignation in the second game on May 4 and the fact he was forced into a draw in the fifth game on Saturday. In the latter game he had the advantage of playing with the white pieces, which allowed him to move first.
Kasparov also introduced an element of controversy tonight when he questioned the origin of several of Deep Blue's moves and pointedly voiced skepticism about the computer's actions in the second game. In that game, Deep Blue made a series of brilliant moves but then failed to anticipate one Kasparov could have made -- but didn't -- to force a tie. Kasparov, who didn't notice the possible move until it was pointed out to him after the game, said: "I still don't understand how the machine couldn't see that."
When directly asked if he was accusing the IBM team of cheating, Kasparov responded: "I suggested that there were things in this match well beyond my understanding. . . . There's probably no way to prove that Deep Blue is making this move or that move."
IBM researcher Chung-Jen Tan, the leader of the Deep Blue team, said the computer received no human assistance during the games. He said his team was "very proud" of the match's outcome.
Kasparov, who has complained that he hasn't been able to study the computer's behavior more fully, today renewed his request that the IBM researchers provide a printout of the computer's log from the previous games, particularly Game 2. On Saturday, IBM officials agreed to place a copy of the logs with a neutral party, but tonight they said they would not release them to Kasparov or the public. Tan said portions might eventually be published in scientific journals.
Some chess experts here suggested that Kasparov's comments about Deep Blue's playing were sour grapes.
"I think it's nonsense," said Patrick Wolff, a grandmaster who watched today's game, held in a skyscraper here. "I don't think there's any evidence that IBM tampered with the machine. He's just making excuses."
Wolff and other chess experts agree that Deep Blue's unflappability and its ability to conceive of and execute moves unanticipated by Kasparov were the keys to its victory. If Kasparov, who admitted to being distraught after the second game, had played that game to a tie, the remaining games might have turned out differently, the experts speculated.
The nine-day match began with a riveting game on May 3 in which Deep Blue, playing black, attempted an aggressive series of moves in mid-game. But Kasparov skillfully used the opportunity to mount a bold counterattack, forcing the computer to concede defeat after nearly four hours.
The next day, however, the computer, showing a finesse never before seen in a chess-playing machine, rallied to force a Kasparov resignation. It was only the second time a computer had defeated the top world champion; the first occurred at the first meeting of Kasparov and Deep Blue last February in Philadelphia, when the computer won the first game but went on to lose the match, 4-2.
In today's game, in which the computer played with the white pieces, Kasparov fell prey to a knight Deep Blue injected into his territory. By using one of his pawns to eliminate the knight, Kasparov opened himself up to attack.
Many observers here were dumbfounded Kasparov could make such a misstep. "Why did he do that?" asked Danny Kopec, a grandmaster and computer science professor who watched the match. "That's what everyone wants to know."
The 1.4-ton supercomputer relies on thousands of lines of complex mathematical equations and logic expressions to find the best move. Deep Blue was designed with the help of several grandmasters to play strategically. It has been programmed not to just make the best immediate move, but to execute a particular series of moves, an enhancement that makes its play appear much less machinelike.
Kasparov received $400,000 from IBM for his participation. IBM said it would devote the $700,000 winner's purse to further computer chess research.
© Copyright 1997 The Washington Post Company